On This Page
> resources > docs > process mining data > Supported Data Formats for Event Logs

Supported Data Formats for Event Logs

Process Intelligence Process Intelligence Available with a Process Intelligence seat or higher. Only seats on this plan or a higher one can use this feature. Compare plans Available with a Process Intelligence seat or higher

File Support for Data Upload

ProcessMind supports many file formats for uploading event logs. All of them can be uploaded through both the application UI and the API.

Text-Based Formats

  • .csv: Comma-separated values. Common and lightweight; most systems (Excel, ERP, and others) can export it. Delimiters are autodetected.
  • .tsv: Tab-separated values. Similar to CSV but uses tabs instead of commas; helpful when data contains commas.
  • .txt: Plain text files. Can be used if structured in a consistent, delimited format.
  • .xes: Standard format for event logs used in process mining tools. Ideal for detailed activity tracking with timestamps and case IDs.

Excel Formats

  • .xls: Excel spreadsheet format (older version). Supported but may load slower than newer formats.
  • .xlsx: Modern Excel format. Widely used and fully supported, often preferred for structured data.
  • .xlsb: Excel Binary Workbook. Loads faster and handles large datasets better than .xlsx, but not as universally supported.

Binary / Columnar Data Formats

  • .parquet: Apache Parquet format. Columnar storage format optimized for analytics workloads. Excellent for large datasets, significantly faster than CSV for big data. Supports incremental (delta) uploads.
  • .orc: Optimized Row Columnar (ORC) format. Another efficient columnar format common in Hadoop/Spark ecosystems. Supports incremental (delta) uploads.
  • .jsonl / .ndjson: JSON Lines format. Each line is a separate JSON object. Great for streaming data exports from APIs and logging systems. Supports incremental (delta) uploads.

Compressed Files

  • .gz: Gzip-compressed files. Any of the above formats can be gzip-compressed (for example, data.csv.gz, data.parquet.gz). ProcessMind automatically detects the inner format and decompresses as needed. Using gzip compression can reduce upload times for large files.

Performance Tip: Use Parquet for Large Datasets

For datasets with millions of rows, consider using Parquet format. Parquet files are columnar, compressed, and include embedded schema information. This means faster uploads, faster queries, and automatic column detection without header row parsing. Parquet also supports incremental (delta) uploads for continuously growing datasets.

General File Structure Requirements

For successful process mining within ProcessMind, your uploaded files (whether they are in Excel, text-based, or XML formats) must adhere to specific structural guidelines. This ensures the app can interpret the data correctly and perform accurate analysis.

1. Header Row (Excel or Text-Based)

  • The file must start with a header row, which should be located in the first line (for example, cell A1 for Excel files or line 1 for CSV, TSV, or TXT files). The header defines the column names and should clearly indicate the type of data in each column (for example, “Case ID,” “Activity,” “Timestamp”).
  • For CSV, TSV, and TXT formats, delimiters and quotes will be autodetected, making it easier to upload your data without needing to specify these settings manually.
  • For Parquet and ORC formats, the column names and data types are read directly from the embedded file schema: no header row is needed.
  • For JSONL/NDJSON formats, column names are derived from the JSON keys in the first line of the file.

2. Minimum Set of Attributes

To support a process mining event log structure, your file must contain, at minimum, the following attributes (columns). These do not need to correspond to the naming of the columns (Case ID, Activity, etc.), but they should contain the content specified.

  • Case ID: This column uniquely identifies each process instance (or case). Every row corresponding to the same process instance must have the same Case ID.
  • Activity: This column should describe the specific activity or event being recorded (for example, “Order Created,” “Payment Processed”).
  • Timestamp: Each activity must be associated with a timestamp that marks the exact time or date the event occurred.
  • Note: The timestamp format will be autodetected as much as possible. Common formats like yyyy-MM-dd HH:mm:ss, MM/dd/yyyy, and others are automatically recognized.
  • Optional Attributes: You may include additional columns to improve your analysis, such as:
  • Resource: Identifies who performed the activity (for example, user, department).
  • Cost: Any costs associated with the activity.
  • Other Custom Data: You can include custom fields that are relevant to your specific process, as long as the required columns are present.

The exact columns depend on the process you are analyzing. Our process improvement guides describe the schema each process needs: the required fields, the activities to capture, and a data template you can fill in.

3. Data Formatting

  • Ensure that your data is consistently formatted across all columns:
  • Timestamps should be in a standard, recognizable format (for example, yyyy-MM-dd HH:mm:ss), though ProcessMind will try to autodetect the date format if it’s different.
  • Avoid blank rows between data entries, as this may disrupt the import process.
  • Ensure that numeric data (for example, costs, durations) is formatted as numbers in Excel, or correctly formatted in text-based files (CSV, TSV, TXT).
  • For CSV, TSV, and TXT formats, ProcessMind will autodetect delimiters (commas, tabs, semicolons, etc.) and handle quoted text automatically.

4. Sheet Selection (For Excel Files Only)

  • ProcessMind automatically processes data from the first sheet in your Excel file (XLS, XLSX, or XLSB), regardless of its name. Ensure that the required event log data is placed on the first sheet, as additional sheets will not be considered during import.

5. XES Format

The .xes (eXtensible Event Stream) format is a standard file type used for event logs in process mining. It captures detailed information about process instances, such as case IDs, activities, timestamps, and other relevant attributes.

It is widely supported by process mining tools and ideal for accurate reconstruction of real-life process behavior.

For technical specifications and full details, visit the official XES standard website.

6. Parquet, ORC, and JSONL Formats

Parquet and ORC are columnar binary formats widely used in data engineering pipelines (Apache Spark, AWS Glue, Databricks, etc.). They include embedded schema information, so ProcessMind automatically detects column names and data types without any manual configuration.

JSONL (JSON Lines, also known as NDJSON) files contain one JSON object per line. Each object represents a single event. The JSON keys become the column names. This format is common when exporting data from APIs, logging systems, or streaming platforms.

All three formats:

7. Tips

Performance Tip: Use XLSB Format for Faster Processing

While all supported Excel formats can be uploaded and processed by ProcessMind, use the XLSB format for Excel files. The XLSB format stores your Excel file in a binary format, which offers significant performance benefits, especially for large datasets. This means faster loading times and quicker processing when compared to XLS or XLSX formats.

For Large Datasets: Use Parquet or ORC

For datasets with millions of rows or automated data pipelines, use Parquet or ORC format instead of Excel or CSV. These columnar formats offer the best compression and query performance. They support incremental loading for continuously growing datasets.

Related Documentation